Struct contents reference from a non-struct array object

(Not matching to prev questions on google).I have struct of struct of arrays. When any struct is empty while concatenating, the above error is displayed.

d=[s1.a.d;s2.b.d;s3.c.d]

when any struct a,b,c is empty above error is shown. How to redefine stacking to eliminate the error.

Answer 1

[Guillaume]

"when any struct a,b,c is empty"

If a is empty (i.e. a = []), then it is not a structure, and a.d doesn't mean anything.

What would be the result of the concatenation anyway in this case.

[Mukund]

This line is iterated,hence kept. On other times, that struct could be non-empty. I wanted to know: either 1) how to improve the concatenation, or 2) introduce check conditions. I have tried isempty and isstruct but they are not working as expected.

[Stephen23]

"If a is empty (i.e. a = []), then it is not a structure, and a.d doesn't mean anything."

MATLAB clearly allows empty structures:

>> a = struct('d',{})
a = 
0x0 struct array with fields:
    a
>> isempty(a)
ans =
     1

So the logical that something empty cannot be a structure is incorrect.

"What would be the result of the concatenation anyway in this case."

The a.d syntax always returns a comma-separated list with as many elements as a has (in this case zero):

>> a.d
>>

So the empty structure does exist, and it (correctly) returns a comma-separated list with zero things in it. I would also expect a comma-separated list from a cell array to deliver the same result:

>> C = {}
C = 
     {}
>> C{:}
>>

So far no surprises for me. Comma-separated lists with zero members can easily be concatenated without any error:

>> b = struct('d',2);
>> [a.d,b.d]
ans =
     2

Any problems are solely due to the data type, and have nothing to do with whether the array is empty or not. Fix the data class.

[Mukund]

Thanks Stephen, you spotted the mistake accurately.